[QUOTE=Dubslow;280166]I just noticed it records all work completed, not just stuff assigned by the system.[/QUOTE]

Um, no...

Could you explain what you mean?

Wait... I'm not sure. I saw some exponents that I thought I got from Mr. P-1. Let me check, brb.

Edit: Nope, my bad. I had a large batch of P-1 of which three were his. I didn't check the size of the batch, only knew three were his. On my completed page, I saw a lot more P-1 than I remembered reserving, so I thought it must have been those, but further checking says they weren't. I had no clue I'd reserved so many P-1's. Sorry.

My DC's still missing from the workers report... Does the GPU-done (as instead of P95-done) has anything to do with it? Or still training the spider? ("Fetch boy, fetch. Sit. Shake pedipalp. Wiggle opisthosoma... Good boy...")

[QUOTE=chalsall;280119][URL="http://gpu.mersenne.info/reports/workers/"]Workers' Progress[/URL] and [URL="http://gpu.mersenne.info/reports/overall/"]Overall System Progress[/URL], plus individuals' "Completed" report.

Obviously the Overall report needs to be reworked to show the number of attempts for DC (and, soon LL). The DC row was simply put there to show that the system is picking them up.

The Overall report needs some TLC anyway, in order to show the number of LL runs (and the associated GHz Days) saved because of the TF and P-1 work which found factors.

Please note that I'm still "training my spider" ("Fetch boy, fetch. Sit. Shake paw. Good boy..."), so the completed DCs are not yet picked up "real time". But they will be soon....[/QUOTE]

Question on the website... On the View Assignments page, would it be a problem to allow an option for ordering by date? I was cutting and pasting from my 2 results files when I made a mistake and somehow copied instead of deleting so when I pasted I had old data and the new stuff is now gone. If I could sort by date, then in a couple of days I could figure out which exponents I fubared and rerun them.

[QUOTE=James Heinrich;279985]i7-920 @ 3.2GHz. On 3 cores I get 0.007s iterations. On 4 cores I get 0.006 :ermm:[/QUOTE]

Looking back, I see you got 0.021, which I'm guessing was on a single core, the 2400 gets 0.016 on a single, haven't tried to bump it up.

[QUOTE=bcp19;280181]Looking back, I see you got 0.021, which I'm guessing was on a single core, the 2400 gets 0.016 on a single, haven't tried to bump it up.[/QUOTE]If you do have a full benchmark, send it to me (PM, email) and I'll put it on my [url=http://mersenne-aries.sili.net/bench.php]benchmarks page[/url] (which automatically also puts it on the [url=http://mersenne-aries.sili.net/throughput.php?cpu1=Intel%28R%29+Core%28TM%29+i7+CPU+920+%40+2.67GHz|256|8192&mhz1=3200]throughput page[/url] with graphs and such.

I put the GPU into it, and let the 25162937 finish. Residue matched. Now let the anonymous guy make the triple check. I know it is not fair for him, and I am quite sorry. But I got angry with P95 :furious:.

What is the better path to go for DC and LL assignments taken from GPU-2-72? I mean, do I have to put them into P95? (to confirm the key and take the ownership?). Will they be unreserved if I just do nothing, and keep them as anonymous, DC/LL them with CudaLucas, and report the residue without the key?

Other better idea?

Something is wrong with my CudaLucas, the time increases instead of decreasing....

[CODE]
e:\<snip\>C4 -c100000 -t -D1 29006071
CUDALucas: Could not find a checkpoint file to resume from
Iteration 100000 M( 29006071 )C, <snip>, ETA 29:16:10)
Iteration 200000 M( 29006071 )C, <snip>, ETA 30:16:07)
Iteration 300000 M( 29006071 )C, <snip>, ETA 31:44:39)
Iteration 400000 M( 29006071 )C, <snip>, ETA 32:06:54)
Iteration 500000 M( 29006071 )C, <snip>, ETA 32:31:03)
Iteration 600000 M( 29006071 )C, <snip>, ETA 32:33:38)
Iteration 700000 M( 29006071 )C, <snip>, ETA 33:03:41)
Iteration 800000 M( 29006071 )C, <snip>, ETA 33:13:00)
[/CODE]

[SPOILER]this post is a joke, I was launching other tasks in between, that took some juice from the CPU/GPU, but it was fun to see it, when I looked to the times I was crossed for few seconds until I realized what was going on[/SPOILER]

[QUOTE=LaurV;280224]I put the GPU into it, and let the 25162937 finish. Residue matched. Now let the anonymous guy make the triple check. I know it is not fair for him, and I am quite sorry. But I got angry with P95 :furious:.[/QUOTE]

But this was assigned to you as a TF (67 -> 69), not a DC... (And this TF work was done: Assigned 2011-11-22 11:34:09, Completed 2011-11-23 00:33:11).

[QUOTE=LaurV;280224]What is the better path to go for DC and LL assignments taken from GPU-2-72? I mean, do I have to put them into P95? (to confirm the key and take the ownership?).[/QUOTE]

Make sure they're actually DC and LL assignments; don't do DC/LL work on DC/LL [B][I][U]Trial Factoring[/U][/I][/B] assignments. And while it is good if you put them in P95 to take ownership (and also to let PrimeNet know that someone's actually working it), it is not fundamentally required.

[QUOTE=LaurV;280224]Will they be unreserved if I just do nothing, and keep them as anonymous, DC/LL them with CudaLucas, and report the residue without the key?[/QUOTE]

Nope. Spidy will detect that you've completed the work, award you your credit, and then unassign the anonymous registration. I realized that this would be a path some would take.

Who's a good spider???
 

OK, just so you know, Spidy has been let off it's leash with regards to detecting and awarding GHzDays credit for completed LL and DC work.

Completed work will be detected at approximately 35 minutes after each hour, and the associated reports will automatically be updated (and the candidate unreserved if appropriate).

I think that's one of the exponents that Prime95 mysteriously unreserved for no reason. So he did have a DC assignment (not TF) at some point...

[QUOTE=chalsall;280230]But this was assigned to you as a TF (67 -> 69), not a DC... (And this TF work was done: Assigned 2011-11-22 11:34:09, Completed 2011-11-23 00:33:11).

Make sure they're actually DC and LL assignments; don't do DC/LL work on DC/LL [B][I][U]Trial Factoring[/U][/I][/B] assignments. And while it is good if you put them in P95 to take ownership (and also to let PrimeNet know that someone's actually working it), it is not fundamentally required.

Nope. Spidy will detect that you've completed the work, award you your credit, and then unassign the anonymous registration. I realized that this would be a path some would take.[/QUOTE]

That exponent was FIRST TIME assigned to me to TF, which I done long ago. Then it was SECOND TIME assigned to me, for Double Check (NOT TF) with a proper assigment key. See [URL="http://www.mersenneforum.org/showpost.php?p=280082&postcount=270"]my post #270[/URL] above.

And what do you want to say with "Nope." ?? Should I take this path, or not? Should I unreserve the exponent or should I not, after DC-ing it with CudaLucas, if it is still reserved as anonymous? I want to avoid using P95 for expos got from GPU2-72, exactly for the reasons explained in post 270.

[QUOTE=LaurV;280340]That exponent was FIRST TIME assigned to me to TF, which I done long ago. Then it was SECOND TIME assigned to me, for Double Check (NOT TF) with a proper assigment key. See [URL="http://www.mersenneforum.org/showpost.php?p=280082&postcount=270"]my post #270[/URL] above.[/QUOTE]

Yes, but, as you said in post #270 above, Prime95 unreserved it. Thus you no longer "owned" it, and should not have completed the DC on it. This has simply resulted in (possible) duplication of effort (as you, again, said in #270).

[QUOTE=LaurV;280340]And what do you want to say with "Nope." ?? Should I take this path, or not? Should I unreserve the exponent or should I not, after DC-ing it with CudaLucas, if it is still reserved as anonymous? I want to avoid using P95 for expos got from GPU2-72, exactly for the reasons explained in post 270.[/QUOTE]

I thought the rest of the paragraph explained it clearly. But to be extra explicit, do [B][I][U]not[/U][/I][/B] unresearve the exponent from your G72 assignments page. This system will pick up that you've completed the work (assuming you're logged in when you submit the results manually) and unassign it automatically within an hour.

Ok. Now is clear.

And to clarify about my mysterious exponent: that was not a complain against GPU272. It was just my anger leashed out, against P95's mysterious** ways, and the "anonymous" who "stolen" my exponent :smile: I know I duplicated (triplicated) the work, but that was my way of "showing off" in his face :P

edit: I said "mysterious", because it was so. It can not be related to "days to unreserve" or other thing like that, as it only happened to worker 2, but not to the other workers. Each worker had 2 (two) DC exponents to crunch, first one from GPU272 and second one from Primenet. I did this specially, I requested work from Primenet to have one exponent in each queue, because I was reading "stories" on the forum about P95 "mysterious unreserving" work when it is "too much to do". So, I said in my mind, I will take 4 expos from GPU272, and INSERT one in each queue, BEFORE the one got from Primenet. My logic was that in such case, if P95 considers that I have "too much work to do", it will unreserve the expos on the "tail" of the queue, which had no work done yet (therefore the one from Primenet) and not the expos in the "front" of the queue, which had some iterations done already. I intended to have the "end of the queue" occupied by Primenet expos, for each worker. But he misteriously decided to unreserve BOTH exponents from worker 2, but keep ALL 6 exponents that were in the queues for workers 1, 3, and 4 (which are still active, P95 is crunching them, first three (number 1 in each queue) are GPU272 expos, last three (number 2 in each queue) are Primenet expos, in the range 29M. Additionally, he got a brand-new expo (29M) from Primenet, which is crunched currently on worker 2. So, why for the other 3 queues the work-to-do was not "too much"?!?!?! Mystery!! I will let them finish in P95, and do not want to touch the worktodo file anymore. Therefore, I want to DC future assignments from GPU272 with CudaLucas, without using P95 to "take ownership", send the results as LaurV, then your spider should unreserve the (still anonymous) assignment. That is all.

Ok, now a small suggestion:

The "overall system progress" report page should include a new column saying "exponents cleared by DC/LL" or equivalent, after the "factors found column", OR this information should be shown in the "factors found" column, with a small red asterisk (because in this case, no factor is found, but in either case, - factor found, or DC/LL completed - the work counts as "the exponent was cleared" - either by the factor which it was found, or by the completion of DC/LL). And in this case, the column "Average blabla per factor" could be modified in "average blabla per factor or expo cleared/completed", or whatever, and instead of N/A, it must show the real figure (it would be an average of whatever credit Primenet gives for each DC/LL). Then the report would be more reliable and accurate. One who is looking there first time and finds "0" and "N/A" wonders "what in the hack that means, and why do these guys work in vain here?". In the case when no column is added, the "factors found" title should be changed in "exponents cleared/completed", because we know that the "clearance" means "factors" for TF things, and it means "test done" for DC. The "completed" part refers to LL, as the "clearance" is not exact when the new row for LL completed will appear, because then, the exponent is not really "cleared". It still have to wait for DC completion... But anyhow, the "0" and the "N/A" look bad there...

[QUOTE=LaurV;280357]And to clarify about my mysterious exponent: that was not a complain against GPU272. It was just my anger leashed out, against P95's mysterious** ways, and the "anonymous" who "stolen" my exponent :smile: I know I duplicated (triplicated) the work, but that was my way of "showing off" in his face :P[/QUOTE]

I know you're relatively new around here, so please take the following as friendly advice from a long term GIMPSer...

You didn't duplicate the work, you caused another user to possibly duplicate it. Since it is an anonymous account, it's possible (but not certain) that the work (legitimately) assigned to them may not actually ever be completed, and thus no harm (may have been) done.

But what you did is very similar to what is known around here as "poaching", and is considered bad "PrimeNetiquette". I know you felt you owned that assignment. But as soon as Prime95 hands it back to PrimeNet (for whatever reasonable or mysterious reason), it is no longer yours.

You hadn't yet done any work on it, had you? You should have simply requested another one and continued on -- there's lots of work to go around.... :smile:

[QUOTE=chalsall;280398]I know you're relatively new around here, so please take the following as friendly advice from a long term GIMPSer...
[/QUOTE]
Well, this is not quite exact... I am new to the "GPU owners" class, and somehow new to the theoretical domain. Being GIMPSter since the beginnings (about 10 years). There was a time when I had 40 computers involved into the project, and I was in top 100 list for years. The interest went down after the 10M-digit prize was awarded, I got tired to run around the building to maintain them, and I only kept the stuff around me (office, home). They slowly died one by one, with new OS re-installations, upgrades, renew the hardware farm in the company, etc. The interest revived since I joined the forum. But thanks for the advice, anyhow :D

[QUOTE=LaurV;280362]Ok, now a small suggestion:
[/QUOTE]
@chalsall: I saw you took this suggestion on the hard way, you just deleted the two rows at all from the report :rant:

Also within two weeks I'll need an LL column in the Worker's Progress window :)

[QUOTE=Dubslow;280494]Also within two weeks I'll need an LL column in the Worker's Progress window :)[/QUOTE]
As I understood, that is already implemented, but waiting for some data (same as it happened with the DC column).

Oh. I'll be happy to oblige. Shall we start a 'pot'?

"Pot"?? All resources busy for the next 40 hours or so... Primenet LL-front assignment (one expo) on one card - about 38 hours ETA, and GPU-2-72 (two expos) DC assignment on second card (ETA about 40 hours together). But if you like, I would stop the current LL when I arrive home this evening and take one LL assignment from GPU272.

Generally I avoided to take "first LL" work from GPU-2-72, because [B]TF-ing at LL front[/B] seems more profitable then LL-ing, about 3 to 5 times more profitable, according with the overall progress table. One can find a factor every two-three days, and saves 2 LL-s which would take 10-14 days.

Contrary to [B]TF-ing at DC-front[/B], which seems less profitable then DC-ing, about half of the benefit, as I said in my former posts in this thread, and it can be seen in the table, now after we collected enough data. One can find a DC-front factor every two days, but he would need one day to DC the exponent and get rid of it. Of course, finding factors is cool...

But I would really suggest TF-only for LL-front, and DC-only for DC-front, for the time being. And that is what I am doing. No TF at DC front, and no LL at LL-front. This to be clear :D

Until the overall tables would look differently... Just my tuppence.

Oops. Turns out my first GPU272 exponents won't be done until December... Unless I put them ahead...
Edit: I've now done that, if only to start two exponents at exactly the same time to see which one completes first, and how far apart they are when they finish. It'll be almost exactly 15 days.
(@LaurV: I do do TF work, on my GPU. I just happen to have cores left over for P95, which primarily run LL tests.)

I have just checked the Worker´s Progress table, and noted that GPU to 72 has become a Holy Project: God Himself did care to join! :smile: (prime95 appears on the list of workers). But so far God didn´t create anything... no results showing up yet.

[QUOTE=lycorn;280548]I have just checked the Worker´s Progress table, and noted that GPU to 72 has become a Holy Project: God Himself did care to join! :smile: (prime95 appears on the list of workers). But so far God didn´t create anything... no results showing up yet.[/QUOTE]

Gee, there is also the Invisible Man! :smile:

Luigi

Hmm.. just became visible (or else vanished from the list). :wink:

[QUOTE=lycorn;280548]I have just checked the Worker´s Progress table, and noted that GPU to 72 has become a Holy Project: God Himself did care to join! :smile: (prime95 appears on the list of workers). But so far God didn´t create anything... no results showing up yet.[/QUOTE]
Could be LL or DC

@chalsall: How about a "total" line for the two tables with "cost per factor found?" Like add all in the "Run" and "F" columns, and use the same formula as in the table to compute the gigs-days. A general total for each table and a general total for all the project would be welcomed too, however this will imply a third small table (with two rows and a total row) beside/under the two tables already there, or, if you don't like the idea with the third table, then a new column on the right of each of the two tables would suffice too, and sum them horizontally. Then the lower-right cells of the tables will show the general total.

[QUOTE=LaurV;281829]@chalsall: How about a "total" line for the two tables with "cost per factor found?" Like add all in the "Run" and "F" columns, and use the same formula as in the table to compute the gigs-days. A general total for each table and a general total for all the project would be welcomed too, however this will imply a third small table (with two rows and a total row) beside/under the two tables already there, or, if you don't like the idea with the third table, then a new column on the right of each of the two tables would suffice too, and sum them horizontally. Then the lower-right cells of the tables will show the general total.[/QUOTE]

I was also thinking a total line for each table would be a nice addition.

[QUOTE=Chuck;281850]I was also thinking a total line for each table would be a nice addition.[/QUOTE]

Your wish is my command.... :smile:

[QUOTE=LaurV;281829]A general total for each table and a general total for all the project would be welcomed too, however this will imply a third small table (with two rows and a total row) beside/under the two tables already there, or, if you don't like the idea with the third table, then a new column on the right of each of the two tables would suffice too, and sum them horizontally. Then the lower-right cells of the tables will show the general total.[/QUOTE]

That summary is already available on the Overall System Progress report. The whole reason of producing the Cost per Factor Found report was to break down the cost to a higher granularity, as the average across all bit levels doesn't really make sense as each level costs twice as much as the previous, with a slightly smaller chance of finding a factor.

Additionally, an overall average really doesn't make sense as each range has a different cost; the larger the exponent, the less time it takes to trial factor. Within say a 10M range this isn't hugely noticeable, but between say 24M and 45M it definitely is.

[QUOTE=chalsall;281981]Your wish is my command.... :smile:[/QUOTE]

I want a Sam Adams Octoberfest...

Very minor
 

Number of factors is incorrect if P-1 does a [URL="http://www.mersenne.org/report_exponent/?exp_lo=51901393&exp_hi=51901393&B1=Get+status"]double find[/URL]...
I can live with this...
BTW, the submission spider is nice! Thanks.

P-1 assignment w/TF to 71
 

I'm pretty sure that something similar has been kicked around before, but I have not been able to track it down. Sorry for: :deadhorse:

I currently have a number of P-1's which have only been taken to 71 bits. At the time, there were no 72 bit exponents available. Now, I am wondering if I can legitimately take these to 72 first, and then run P-1.

[QUOTE=kladner;284112]I currently have a number of P-1's which have only been taken to 71 bits. At the time, there were no 72 bit exponents available. Now, I am wondering if I can legitimately take these to 72 first, and then run P-1.[/QUOTE]

Certainly you can.

However, please note that if these were assigned to you by G72 then you won't get the TF credit for the work from 71 to 72 on G72 because the system isn't watching for this and it would be a pain (and extra load on PrimeNet) to do so.

You will, of course, get the credit on PrimeNet.

[QUOTE=chalsall;284114]Certainly you can.

However, please note that if these were assigned to you by G72 then you won't get the TF credit for the work from 71 to 72 on G72 because the system isn't watching for this and it would be a pain (and extra load on PrimeNet) to do so.

You will, of course, get the credit on PrimeNet.[/QUOTE]

Thanks very much, chalsall. It is good to know.

[QUOTE=chalsall;278946]Cute idea. :smile:

I have created the team "GPU to 72" if anyone wants to join....[/QUOTE]

If I join the team, do my GHz-Days stop going to me?

It's a question of feeding my ego and the Top Producers list. As ktony said, it's my sport's score page. :smile:

[QUOTE=oswald;284661]If I join the team, do my GHz-Days stop going to me?

It's a question of feeding my ego and the Top Producers list. As ktony said, it's my sport's score page. :smile:[/QUOTE]

No you will still get your credits. I am on the team and also about 40th overall and 20-something in TF on the top producers list.

Chuck,
Thanks, Joining now. :smile:

Sweet! The sad part is, everybody on the team, running for the last 2 months, still hasn't equaled the throughput of Xyzzy or nucleon, individually!

Welcome, oswald!

As to equalling Xyzzy or nucleon......not everyone has the hardware or the power capacity to match those titans.

[QUOTE=kladner;284683]Welcome, oswald!

As to equalling Xyzzy or nucleon......not everyone has the hardware or the power capacity to match those titans.[/QUOTE]

But between 6 of us, and now 7? Maybe the 7th will push our throughput to match Xyzzy. (7+ to 1....!)